Digital twinborn modeling and predictive analysis system for oil and gas well

Abstract

The invention relates to the technical field of prediction algorithms, in particular to an oil and gas well digital twinborn modeling and prediction analysis system which comprises an underground dynamic data analysis module for extracting variables based on pressure, flow velocity and temperature data, performing data association calculation, analyzing the pressure change rate and the flow velocity trend, calculating the temperature fluctuation range and establishing time period variable analysis; and generating a dynamic parameter mode. According to the invention, through an autoregressive integral moving average model, the dynamic data of each time period is integrated and smoothed, the long-term trend of the data is captured, the subtle change in the production fluctuation is identified, the accurate future yield fluctuation trend is generated, a density clustering algorithm is adopted, key variables are analyzed, and the change mode of abnormal fluctuation is extracted. According to the method, potential fault types are screened, similar data points are classified through the density relation between data, dependence on parameter setting is reduced, abnormity is recognized more accurately under the unsupervised condition, and the fault recognition capability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of prediction algorithms, and in particular to an oil and gas well digital twin modeling and prediction analysis system. Background Art

[0002] The field of predictive algorithm technology focuses on predicting future states or events based on historical data, real-time data or simulation data through computer systems, algorithms, machine learning and artificial intelligence technologies, in order to optimize the exploitation of oil and gas resources, improve production efficiency, reduce potential risks, and make scientific decisions in complex working conditions.

[0003] The purpose of the digital twin modeling and predictive analysis system for oil and gas wells is to establish a highly simulated digital model of oil and gas wells, combine a large amount of production data and environmental data, and predict downhole conditions, such as reservoir changes and pressure changes, so as to optimize mining decisions, improve resource recovery rates, and reduce risks and costs.

[0004] Traditional systems rely on simple linear regression or rule-setting algorithms, which make it difficult to handle complex nonlinear relationships between multi-dimensional variables, resulting in inaccurate predictions of production fluctuations, lack of in-depth analysis of the correlation between multiple variables such as pressure, flow rate, and temperature, and inability to fully capture the complexity of downhole conditions. The lack of automated anomaly detection and analysis methods results in long fault detection response times, missed early warning signals, untimely maintenance of downhole equipment, increased production risks and costs, and waste of resources and safety hazards. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a digital twin modeling and predictive analysis system for oil and gas wells.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A digital twin modeling and prediction analysis system for oil and gas wells comprises: Downhole dynamic data analysis module: Based on the pressure, flow rate and temperature data of downhole equipment, extract variables, perform correlation calculations between data, analyze the pressure change rate and flow rate trend, calculate the temperature fluctuation range, establish variable analysis for multiple time periods, and generate dynamic parameter change patterns; Digital twin modeling generation module: Based on the dynamic parameter change mode, the autoregressive integral moving average model is adopted to integrate the dynamic data of each time period, and the global data model of the oil and gas well is constructed by comprehensive analysis of the production history data and the real-time data, so as to form a virtualized real-time mapping of the oil and gas well and generate the digital twin modeling of the oil and gas well; Production fluctuation prediction module: Based on the oil and gas well digital twin modeling, calculate future production fluctuations, extract the change values ​​of pressure and flow rate within the production cycle, analyze historical and future production data, screen out data intervals with large changes and judge their trends, and generate future production fluctuation trends; Abnormal fluctuation monitoring module: based on the future production fluctuation trend, compare the deviation between actual production and predicted value, calculate the deviation of each variable, extract abnormal points, judge the trend deviation of pressure and flow rate changes, and generate production abnormal fluctuation results in combination with the abnormal fluctuation time period; Multi-dimensional fault identification module: Based on the abnormal production fluctuation results, a density clustering algorithm is used to analyze key variables, extract key variables in historical data, analyze the change pattern of abnormal fluctuations, combine the historical trends of flow rate and pressure, screen possible fault types, and generate fault pattern identification results; Correlation pattern analysis module: Based on the fault pattern identification results, analyze the equipment operating conditions and geological conditions, extract the correlation of fault patterns, determine the correlation between fault types and different environments, calculate the occurrence frequency of each fault mode, and generate multi-layer correlation pattern analysis results.

[0007] As a further scheme of the present invention, the dynamic parameter change mode includes a pressure change rate mode, a flow rate change trend mode, and a temperature fluctuation range; the future production fluctuation trend includes historical production analysis, future production fluctuation calculation, and key data interval screening; the abnormal production fluctuation results include abnormal pressure fluctuations, abnormal flow rate fluctuations, and abnormal fluctuation time periods; the fault mode identification results include pressure fault mode, flow rate fault mode, and historical trend fault type screening.

[0008] As a further solution of the present invention, the downhole dynamic data analysis module includes a pressure analysis submodule, a flow rate analysis submodule, and a temperature analysis submodule, wherein: Pressure analysis submodule: Based on the pressure data of downhole equipment, extract the pressure change value, segment the pressure changes in different time periods, calculate the pressure fluctuation rate and filter the fluctuation range, analyze the change trend of each period, and generate the pressure change pattern; Flow rate analysis submodule: based on the pressure change pattern, extract flow rate data, compare the trend of time and flow rate change, calculate the synchronization of flow rate and pressure fluctuation, screen the time interval with significant flow rate change, compare the fluctuation trend, and generate the flow rate change pattern; Temperature analysis submodule: Based on the flow rate change pattern, extract temperature data, calculate the temperature fluctuation range in each time period, screen the abnormal temperature fluctuation range, combine the flow rate and pressure changes, analyze the correlation between the parameters, and generate a dynamic parameter change pattern.

[0009] As a further solution of the present invention, the digital twin modeling generation module includes a historical data integration submodule, a real-time data processing submodule, and a virtual mapping generation submodule, wherein: Historical data integration submodule: Based on the dynamic parameter change mode, the autoregressive integral moving average model is used to extract the historical production data of the equipment, integrate the data of different production cycles, conduct trend change analysis, summarize the intervals with a long time span, and generate a historical data model; Real-time data processing submodule: based on the historical data model, extract real-time data, compare the real-time changes in each time period, summarize the data fluctuations of real-time pressure, flow rate and temperature in sections, establish the corresponding relationship between the time axis and data changes, and generate a real-time data model; Virtual mapping generation submodule: Based on the real-time data model, historical and real-time data are integrated, the dynamic changes of downhole equipment are virtually mapped, a real-time mapping model of global data is established, and digital twin modeling of oil and gas wells is generated.

[0010] As a further solution of the present invention, the autoregressive integrated moving average model is according to the formula:

[0011] in: The device production data for the current time point, is a constant term, is the autoregression coefficient, The lag time point Equipment production data, is the sliding average coefficient, The lag time point The error term, is the random error term at the current time point, is the production rate coefficient, is the production rate at the current time point, is the device temperature correction coefficient, The device temperature at the current time point.

[0012] As a further solution of the present invention, the production fluctuation prediction module includes a historical data extraction submodule, a fluctuation trend calculation submodule, and a data interval screening submodule, wherein: Historical data extraction submodule: Based on the oil and gas well digital twin modeling, extract the pressure and flow rate data of downhole equipment, analyze the changes in the production cycle, screen the pressure and flow rate values ​​in the key period, integrate the data of different stages and store them by category, and generate a historical production data set; Fluctuation trend calculation submodule: based on the historical production data set, compare the pressure and flow rate changes in different time periods, calculate the amplitude of the fluctuation trend, process the historical data of each time period, determine the change trend of pressure and flow rate, combine the real-time data to perform an extended analysis of the historical trend, and generate the fluctuation trend analysis results; Data interval screening submodule: Based on the fluctuation trend analysis results, screen out the time intervals with large pressure and flow rate changes, extract the values ​​of different fluctuation intervals, analyze the fluctuation amplitude and duration, classify and mark the abnormal amplitude intervals and make trend judgments to generate future production fluctuation trends.

[0013] As a further solution of the present invention, the abnormal fluctuation monitoring module includes a prediction deviation calculation submodule, an abnormal point extraction submodule, and a fluctuation trend deviation judgment submodule, wherein: Prediction deviation calculation submodule: Based on the future production fluctuation trend, the actual data of current production pressure and flow rate are extracted, compared with the predicted values, the deviation value in each time period is calculated, the deviation of different variables is classified, and an overall deviation data set is formed to generate a production prediction deviation data set; Abnormal point extraction submodule: based on the production prediction deviation data set, screen out data points with large deviations, extract abnormal changes in pressure and flow rate, classify and summarize different abnormal types, extract the most critical abnormal points in the production process, establish an abnormal data table, and generate a key abnormal data set; Fluctuation trend deviation judgment submodule: Based on the key abnormal data set, the fluctuation trend deviation of pressure and flow rate is judged, and the data of abnormal fluctuation period is combined to analyze the change of deviation value in different time periods, summarize the characteristics of deviation fluctuation, and generate production abnormal fluctuation results.

[0014] As a further solution of the present invention, the multi-dimensional fault identification module includes a key variable analysis submodule, a fluctuation mode screening submodule, and a fault type screening submodule, wherein: Key variable analysis submodule: Based on the abnormal production fluctuation results, a density clustering algorithm is used to extract pressure and flow rate data, analyze the pressure change amplitude in different time periods, calculate the increase and decrease of flow rate, classify related variables, screen out key data points, and generate a key variable set; Fluctuation pattern screening submodule: based on the key variable set, extract the historical data of key variables, analyze the change pattern of each variable in different time periods, screen out the time periods with large abnormal fluctuations, combine the trends of pressure and flow rate, summarize the fluctuation characteristics of each variable, and generate a fluctuation pattern set; Fault type screening submodule: Based on the fluctuation pattern set and combined with the change history of flow rate and pressure, the fluctuation patterns that may cause equipment failure are screened out, the fluctuation types are classified, the type and characteristics of each fault mode are determined, and the fault pattern recognition results are generated.

[0015] As a further solution of the present invention, the density clustering algorithm is according to the formula:

[0016] in: is the improved local density value, is the number of data points, is the target data point, For the data points, For the Data points With the target data point The Euclidean distance between For the The pressure variation of each data point is For the The flow velocity variation of each data point is For the The length of the time period where the data point is located is other time units, is the Euclidean distance The weight coefficient of is the pressure variation The weight coefficient of is the flow velocity variation The weight coefficient of The length of the time period The weight coefficient of is the bandwidth parameter of the Gaussian kernel function.

[0017] As a further solution of the present invention, the correlation mode analysis module includes an equipment operating condition analysis submodule, a fault mode correlation submodule, and a fault frequency calculation submodule, wherein: Equipment operating condition analysis submodule: Based on the fault mode recognition results, extract the operating data of the equipment under different environmental conditions, analyze the changes in flow rate and pressure under various environmental conditions, screen out geological conditions and parameters related to equipment operation, classify the operating performance of the equipment, and generate an equipment operating condition set; Fault mode association submodule: based on the equipment operating condition set, extract the equipment fault data, analyze the correlation between each fault and the environmental conditions, screen out the matching between the fault and the geological environment, analyze the possibility of the fault, and generate the fault mode association result; Fault frequency calculation submodule: Based on the fault mode association results, extract the historical occurrence data of each fault mode, analyze the occurrence frequency of various faults under different conditions, classify the time distribution of fault modes, screen out the high-incidence periods of faults, and generate multi-layer association mode analysis results.

[0018] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the dynamic data of each time period is integrated and smoothed through the autoregressive integral moving average model, so that the global data model of the oil and gas well can more accurately reflect the real-time dynamic changes of the oil and gas well, capture the long-term trend of the data, identify subtle changes in production fluctuations, and generate accurate future production fluctuation trends. The density clustering algorithm is used to analyze key variables, extract the change pattern of abnormal fluctuations, and screen potential fault types. Similar data points are classified through the density relationship between data, reducing dependence on parameter settings, more accurately identifying anomalies in an unsupervised situation, and providing multi-dimensional fault modes, thereby improving fault identification capabilities, discovering anomalies in advance during the production process of oil and gas wells, avoiding potential safety hazards and losses, and optimizing mining efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a system flow chart of the present invention; Figure 2 Schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] Embodiment 1 See also Figure 1 The present invention provides a technical solution: a digital twin modeling and prediction analysis system for oil and gas wells includes: Downhole dynamic data analysis module: Based on the pressure, flow rate and temperature data of downhole equipment, extract variables, perform correlation calculations between data, analyze the pressure change rate and flow rate trend, calculate the temperature fluctuation range, establish variable analysis for multiple time periods, and generate dynamic parameter change patterns; Digital twin modeling generation module: Based on the dynamic parameter change pattern, the autoregressive integral moving average model is adopted to integrate the dynamic data of each time period, and the global data model of the oil and gas well is constructed by comprehensive analysis of production history data and real-time data, forming a virtual real-time mapping of the oil and gas well, and generating the digital twin modeling of the oil and gas well; Production fluctuation prediction module: Based on the digital twin modeling of oil and gas wells, it calculates future production fluctuations, extracts the change values ​​of pressure and flow rate within the production cycle, analyzes historical and future production data, screens out data intervals with large changes and determines their trends, and generates future production fluctuation trends; Abnormal fluctuation monitoring module: Based on the future production fluctuation trend, compare the deviation between actual production and predicted value, calculate the deviation of each variable, extract abnormal points, judge the trend deviation of pressure and flow rate changes, and generate production abnormal fluctuation results in combination with the abnormal fluctuation time period; Multi-dimensional fault identification module: Based on the abnormal production fluctuation results, the density clustering algorithm is used to analyze key variables, extract key variables in historical data, analyze the change pattern of abnormal fluctuations, combine the historical trends of flow rate and pressure, screen possible fault types, and generate fault pattern identification results; Correlation pattern analysis module: Based on the fault pattern identification results, analyze the equipment operating conditions and geological conditions, extract the correlation of fault patterns, determine the correlation between fault types and different environments, calculate the frequency of occurrence of each fault mode, and generate multi-layer correlation pattern analysis results.

[0022] Dynamic parameter change modes include pressure change rate mode, flow rate change trend mode, and temperature fluctuation range. Future production fluctuation trends include historical production analysis, future production fluctuation calculation, and key data interval screening. Abnormal production fluctuation results include abnormal pressure fluctuations, abnormal flow rate fluctuations, and abnormal fluctuation time periods. Fault mode identification results include pressure fault mode, flow rate fault mode, and historical trend fault type screening.

[0023] See also Figure 2 The downhole dynamic data analysis module includes a pressure analysis submodule, a flow rate analysis submodule, and a temperature analysis submodule, among which: Pressure analysis submodule: Based on the pressure data of downhole equipment, extract the pressure change value, segment the pressure changes in different time periods, calculate the pressure fluctuation rate and filter the fluctuation range, analyze the change trend of each period, and generate the pressure change pattern; Flow rate analysis submodule: based on the pressure change pattern, extract flow rate data, compare the trend of time and flow rate change, calculate the synchronization of flow rate and pressure fluctuation, screen the time interval with significant flow rate change, compare the fluctuation trend, and generate the flow rate change pattern; Temperature analysis submodule: Based on the flow rate change pattern, extract temperature data, calculate the temperature fluctuation range in each time period, screen the abnormal temperature fluctuation range, combine the flow rate and pressure changes, analyze the correlation between the parameters, and generate a dynamic parameter change pattern; Pressure analysis submodule: Based on the pressure data of downhole equipment, the pressure change value is extracted by the first-order difference method, and the pressure changes in different time periods are processed in sections. Specifically, the pressure data is cut according to a fixed time window, and the difference between the pressure values ​​of adjacent time periods is calculated. The pressure change rate is calculated using a sliding window, and the pressure change trend is smoothed using a double exponential smoothing method, where the smoothing coefficient α is set to 0.3. The intervals where the pressure fluctuation amplitude exceeds the set threshold are screened out, and the pressure data of different time periods are classified using the judgment condition based on the pressure change threshold to generate a pressure change pattern; Flow rate analysis submodule: Based on the pressure change pattern, the cross-correlation analysis method is used to extract flow rate data, and the cross-correlation function is used to compare the time and flow rate change trends, and the synchronization of flow rate and pressure fluctuations is calculated. The time lag parameter of the cross-correlation function is set to 1 second, and the correlation coefficient between flow rate and pressure fluctuations in each time period is calculated. The time interval with a flow rate change amplitude greater than the set value is screened, and the local weighted regression method is used to fit the flow rate data. The window size of the local weighted regression is 5 data points. The flow rate change rate is combined with the pressure change pattern to generate the flow rate change pattern; Temperature analysis submodule: Based on the flow rate change pattern, discrete Fourier transform is used to extract temperature data, and the temperature data is converted to frequency domain space. The temperature fluctuation amplitude in each time period is calculated, and the frequency resolution is set to 0.01Hz. The frequency interval where the temperature change amplitude exceeds the set threshold is screened out, and the frequency domain data is restored to time domain data using inverse Fourier transform. The flow rate and pressure fluctuation data are combined for collaborative analysis, and the Pearson correlation coefficient is used to calculate the correlation between temperature, flow rate and pressure to generate a dynamic parameter change pattern.

[0024] See also Figure 2 The digital twin modeling generation module includes a historical data integration submodule, a real-time data processing submodule, and a virtual mapping generation submodule, among which: Historical data integration submodule: Based on the dynamic parameter change mode, the autoregressive integral moving average model is used to extract the historical production data of the equipment, integrate the data of different production cycles, conduct trend change analysis, summarize the intervals with a long time span, and generate a historical data model; Real-time data processing submodule: Based on the historical data model, extract real-time data, compare the real-time changes in each period, summarize the data fluctuations of real-time pressure, flow rate, and temperature in sections, establish the corresponding relationship between the time axis and data changes, and generate a real-time data model; Virtual mapping generation submodule: Based on the real-time data model, it integrates historical and real-time data, performs virtual mapping of the dynamic changes of downhole equipment, establishes a real-time mapping model of global data, and generates digital twin modeling of oil and gas wells; Historical data integration submodule: Based on the dynamic parameter change pattern, the autoregressive integral moving average model is used to extract the historical production data of the equipment. The model order parameters are set to p=2, d=1, q=2, and the data of different production cycles are integrated. The least squares method is used to estimate the parameters of the historical data. During the processing, the sliding window method is used to segment the historical data by year, and the window size is set to 365 days. The production trend changes in each time period are calculated, and the model parameters are optimized through repeated iterations. The characteristics of the interval data changes with a longer time span are summarized to generate a historical data model; Real-time data processing submodule: Based on the historical data model, real-time data is extracted, and the exponentially weighted moving average (EWMA) model is used. The attenuation factor λ is set to 0.2, and the real-time changes in each time period are compared. The data fluctuations of real-time pressure, flow rate, and temperature are processed by segmentation. The segmentation time window is set to once per hour. The missing data is inserted on the time axis using the linear interpolation method, and the corresponding relationship between the time axis and the real-time data changes is established. The fluctuation characteristics of the real-time data are analyzed to generate a real-time data model; Virtual mapping generation submodule: Based on the real-time data model, historical and real-time data are integrated, and the Kalman filter algorithm is used to virtually map the dynamic changes of downhole equipment. The process noise covariance matrix of the Kalman filter algorithm is set to Q=0.001, and the measurement noise covariance matrix is ​​set to R=0.01. The state estimation equation is used to combine historical and real-time data for dynamic estimation and update, and a virtual mapping model of global data is established. The digital twin modeling of oil and gas wells is generated by iteratively updating the real-time estimation value.

[0025] Autoregressive integrated moving average model, according to the formula:

[0026] in: The device production data for the current time point, is a constant term, is the autoregression coefficient, The lag time point Equipment production data, is the sliding average coefficient, The lag time point The error term, is the random error term at the current time point, is the production rate coefficient, is the production rate at the current time point, is the device temperature correction coefficient, The device temperature at the current time point; Execution process: First, the production data of the equipment Through the constant As a basic value, it represents the average production level of oil and gas wells under normal operating conditions. After autoregression coefficient The weighted calculation is used to measure the impact of historical data on current data, and the lag order Determines how many historical data points to use, and the sliding average part is passed through the lagged error term Combined with the moving average coefficient Correction, error lag order New production rate parameter used to determine the number of smoothing geometric error points Indicates the current production rate of oil and gas wells, combined with the rate coefficient Perform weighted calculations to reflect the direct impact of rate changes on production data. Rate data is acquired and dynamically adjusted through real-time sensors. The operating temperature of the equipment By temperature correction factor Weighted to correct the impact of temperature changes on equipment production. Temperature data is collected in real time by sensors, and the random error term at the current time point It reflects the unforeseen random fluctuations in oil and gas well production, generates production data prediction results of oil and gas wells at the current point in time, and provides support for the precise analysis of the digital twin model and the prediction of future production trends.

[0027] See also Figure 2 The production fluctuation prediction module includes a historical data extraction submodule, a fluctuation trend calculation submodule, and a data interval screening submodule, among which: Historical data extraction submodule: Based on the digital twin modeling of oil and gas wells, it extracts the pressure and flow rate data of downhole equipment, analyzes the changes in the production cycle, screens the pressure and flow rate values ​​in key periods, integrates the data of different stages and stores them by categories, and generates historical production data sets; Fluctuation trend calculation submodule: Based on the historical production data set, compare the pressure and flow rate changes in different time periods, calculate the amplitude of the fluctuation trend, process the historical data of each period, determine the change trend of pressure and flow rate, combine the real-time data to perform the extension analysis of the historical trend, and generate the fluctuation trend analysis results; Data interval screening submodule: Based on the fluctuation trend analysis results, screen out the time intervals with large pressure and flow rate changes, extract the values ​​of different fluctuation intervals, analyze the fluctuation amplitude and duration, classify and mark the abnormal amplitude intervals, make trend judgments, and generate future production fluctuation trends; Historical data extraction submodule: Based on the digital twin modeling of oil and gas wells, the pressure and flow rate data of downhole equipment are extracted, and the time series segmentation method is used for data analysis. The time window is set to once a month. The pressure and flow rate in each window are statistically processed, and the data points with large fluctuations are screened out using standard deviation calculation. The screened data will be divided according to the production cycle and the pressure and flow rate data of different stages will be integrated. The data will be classified and stored using a binary classification storage structure to generate a historical production data set; Fluctuation trend calculation submodule: Based on the historical production data set, the pressure and flow rate changes in different time periods are compared, and the linear regression model is used to calculate the amplitude of the fluctuation trend. The loss function of the regression model is the least squares method. The historical data of each time period is processed, and the confidence interval of the regression model is set to 95%. The change trend of pressure and flow rate is determined, and the weighted moving average method is used in combination with real-time data to perform an extension analysis of the historical trend and generate the fluctuation trend analysis results; Data interval screening submodule: Based on the fluctuation trend analysis results, the Z-score standardization method is used to screen out the time intervals with large pressure and flow rate changes. The Z-score threshold is set to 2, and the values ​​of different fluctuation intervals are extracted. The sliding window method is used to calculate the fluctuation amplitude and duration. The sliding window size is set to 15 data points, and the abnormal amplitude intervals are classified and marked. The future production fluctuation trend is judged in combination with the change rate trend, and the future production fluctuation trend is generated.

[0028] See also Figure 2 The abnormal fluctuation monitoring module includes a prediction deviation calculation submodule, an abnormal point extraction submodule, and a fluctuation trend deviation judgment submodule, among which: Prediction deviation calculation submodule: Based on the future production fluctuation trend, the actual data of current production pressure and flow rate are extracted, compared with the predicted values, the deviation value in each period is calculated, the deviation of different variables is classified, and the overall deviation data set is formed to generate the production prediction deviation data set; Abnormal point extraction submodule: Based on the production forecast deviation data set, filter out data points with large deviations, extract abnormal changes in pressure and flow rate, classify and summarize different abnormal types, extract the most critical abnormal points in the production process, establish an abnormal data table, and generate a key abnormal data set; Fluctuation trend deviation judgment submodule: Based on the key abnormal data set, it judges the fluctuation trend deviation of pressure and flow rate, combines the data of abnormal fluctuation period, analyzes the change of deviation value in different time periods, summarizes the characteristics of deviation fluctuation, and generates production abnormal fluctuation results; Prediction deviation calculation submodule: Based on the future production fluctuation trend, the actual data of current production pressure and flow rate are extracted, and the mean square error algorithm is used to compare the actual data with the predicted value. Specifically, the sum of square errors of pressure and flow rate in each time period is calculated, and the errors in different time windows are averaged. The time window size is set to 10 minutes, and the errors in each time period are classified. The standard deviation calculation method is used to summarize the deviation of each variable to form an overall deviation data set. Finally, according to the MSE calculation results of each time period, the production prediction deviation data set is generated; Abnormal point extraction submodule: Based on the production forecast deviation data set, the local abnormal factor algorithm is used to screen out data points with large deviations. The neighborhood parameter of the LOF algorithm is set to 20, and the ratio of local density to neighborhood density of each data point is extracted. Data points with large density differences are classified as abnormal data. Abnormal changes in pressure and flow rate are further extracted, and cluster analysis is used to classify and summarize abnormal types. The K-Means algorithm is used in the clustering process, and the K value is set to 3 to divide different types of abnormal data, extract the most critical abnormal points in the production process, establish an abnormal data table, and generate a key abnormal data set; Fluctuation trend deviation judgment submodule: Based on the key abnormal data set, linear regression analysis is used to judge the fluctuation trend deviation of pressure and flow rate. The independent variable of the regression model is set as time, and the dependent variable is the deviation value of pressure and flow rate. Combined with the data in the abnormal fluctuation period, the changes in the deviation values ​​in different time periods are analyzed, and the positive and negative signs of the regression coefficients are used to judge the fluctuation trend, summarize the characteristics of the deviation fluctuation, and generate the production abnormal fluctuation results.

[0029] See also Figure 2 The multi-dimensional fault identification module includes a key variable analysis submodule, a fluctuation pattern screening submodule, and a fault type screening submodule, among which: Key variable analysis submodule: Based on the abnormal production fluctuation results, the density clustering algorithm is used to extract pressure and flow rate data, analyze the pressure change amplitude in different time periods, calculate the increase and decrease of flow rate, classify related variables, screen out key data points, and generate key variable sets; Fluctuation pattern screening submodule: Based on the key variable set, extract the historical data of key variables, analyze the change pattern of each variable in different time periods, screen out the time periods with large abnormal fluctuations, combine the trends of pressure and flow rate, summarize the fluctuation characteristics of each variable, and generate a fluctuation pattern set; Fault type screening submodule: Based on the fluctuation pattern set and combined with the change history of flow rate and pressure, the fluctuation patterns that may cause equipment failure are screened out, the fluctuation types are classified, the type and characteristics of each fault mode are determined, and the fault pattern recognition results are generated; Key variable analysis submodule: Based on the abnormal production fluctuation results, the density clustering algorithm is used to extract pressure and flow rate data. The minimum sample number is set to 5, the neighborhood radius parameter is set to 0.5, the local density of the data points is calculated, and the areas with higher density are identified as clusters. The points with lower density are marked as abnormal points. The pressure change amplitude in different time periods is analyzed, and the pressure change rate in each time period is calculated using the first-order difference method. When processing flow rate data, the sliding average algorithm is used to calculate the increase and decrease of flow rate. The relevant variables are classified and analyzed, and the key data points with the largest local density are screened out to generate a key variable set. Fluctuation pattern screening submodule: Based on the key variable set, the historical data of key variables are extracted, and the sliding window method is used. The window size is set to 24 hours. The change pattern of each variable in different windows is gradually analyzed. The peak and valley values ​​of the variables are screened out by combining the local extreme value detection algorithm. The time period with large abnormal fluctuations is selected, and the slope of the pressure and flow rate trend is calculated by linear regression analysis. The fluctuation characteristics of each variable are summarized to generate a fluctuation pattern set. Fault type screening submodule: Based on the fluctuation pattern set and combined with the historical changes of flow rate and pressure, the support vector machine classification algorithm is used to screen out the fluctuation patterns that may cause equipment failures. The radial basis kernel function is used with the parameter C set to 1 and γ set to 0.1 to classify each fluctuation pattern and categorize the fluctuation type. The classification results are used to determine the type and characteristics of each fault mode and generate fault mode recognition results.

[0030] Density clustering algorithm, according to the formula:

[0031] in: is the improved local density value, is the number of data points, is the target data point, For the data points, For the Data points With the target data point The Euclidean distance between For the The pressure variation of each data point is For the The flow velocity variation of each data point is For the The length of the time period where the data point is located is other time units, is the Euclidean distance The weight coefficient of is the pressure variation The weight coefficient of is the flow velocity variation The weight coefficient of The length of the time period The weight coefficient of is the bandwidth parameter of the Gaussian kernel function; Execution process: First, select the target data points for oil and gas well production data and , and calculate the data points and The Euclidean distance between , reflecting the difference of data points in the production parameter space. The larger the distance, the lower the similarity between data points. Combined with the pressure change amplitude in the oil and gas well and flow rate variation , respectively and As a weight coefficient, adjust the impact in the overall density calculation to ensure the accuracy of the analysis of production anomalies caused by pressure and flow rate changes, and the length of the time period As an important parameter of time factor, the weight coefficient Adjust to measure the impact of different time periods on density calculation. Weight coefficient , , and The value of is adjusted by regression analysis and gradient descent method on the historical data of oil and gas wells to ensure the adaptability and accuracy of the formula under different working conditions. Control the distribution range of local density and generate target data points The local density value of .

[0032] See also Figure 2 The correlation mode analysis module includes a device operation condition analysis submodule, a fault mode correlation submodule, and a fault frequency calculation submodule, wherein: Equipment operating condition analysis submodule: Based on the fault mode recognition results, extract the equipment operating data under different environmental conditions, analyze the changes in flow rate and pressure under various environmental conditions, screen out the geological conditions and parameters related to equipment operation, classify the equipment's operating performance, and generate an equipment operating condition set; Fault mode association submodule: Based on the equipment operating condition set, extract the equipment fault data, analyze the correlation between each fault and the environmental conditions, screen out the matching between the fault and the geological environment, analyze the possibility of the fault, and generate the fault mode association results; Fault frequency calculation submodule: Based on the fault mode association results, extract the historical occurrence data of each fault mode, analyze the occurrence frequency of various faults under different conditions, classify the time distribution of fault modes, screen out the high-incidence periods of faults, and generate multi-layer association mode analysis results; Equipment operating condition analysis submodule: Based on the fault mode recognition results, the operating data of the equipment under different environmental conditions are extracted, and the changes of flow rate and pressure under different environmental conditions are analyzed using a multivariate linear regression model. The independent variables of the regression model are set as environmental parameters, and the dependent variables are flow rate and pressure. The least squares method is used to calculate the regression coefficients under various environmental conditions, and the operating performance of the equipment under various conditions is classified. The main geological conditions and parameters related to the equipment operation are screened out using principal component analysis. The number of principal components is set to 3 to generate a set of equipment operating conditions. Fault mode association submodule: Based on the equipment operating condition set, the equipment fault data is extracted, and the Pearson correlation coefficient calculation method is used to analyze the correlation between each fault and the environmental conditions. The Pearson coefficient threshold is set to 0.7, and the significant matching between the fault and different geological environmental conditions is screened out. The multivariate discriminant analysis method is used to evaluate the possibility of each fault. The dependent variable in the discriminant analysis model is set as the fault type, and the independent variable is the environmental parameter of the equipment. The matching of different faults under various environmental conditions is classified to generate the fault mode association results; Fault frequency calculation submodule: Based on the fault mode association results, the historical occurrence data of each fault mode is extracted, and the frequency distribution analysis of the fault data is performed using the histogram analysis method. The time interval is set to 1 year, and the occurrence frequency of various types of faults under different geological and equipment conditions is analyzed. The time distribution of various types of fault modes is classified, and the weighted moving average algorithm is used to screen out the high-incidence periods of faults. The weight attenuation coefficient is set to 0.8 to generate multi-layer association pattern analysis results.

[0033] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A digital twin modeling and prediction analysis system for oil and gas wells, It is characterized in that The system comprises: Downhole dynamic data analysis module: Based on the pressure, flow rate and temperature data of downhole equipment, extract variables, perform correlation calculations between data, analyze the pressure change rate and flow rate trend, calculate the temperature fluctuation range, establish variable analysis for multiple time periods, and generate dynamic parameter change patterns; Digital twin modeling generation module: Based on the dynamic parameter change mode, the autoregressive integral moving average model is adopted to integrate the dynamic data of each time period, and the global data model of the oil and gas well is constructed by comprehensive analysis of the production history data and the real-time data, so as to form a virtualized real-time mapping of the oil and gas well and generate the digital twin modeling of the oil and gas well; Production fluctuation prediction module: Based on the oil and gas well digital twin modeling, calculate future production fluctuations, extract the change values ​​of pressure and flow rate within the production cycle, analyze historical and future production data, screen out data intervals with large changes and judge their trends, and generate future production fluctuation trends; Abnormal fluctuation monitoring module: based on the future production fluctuation trend, compare the deviation between actual production and predicted value, calculate the deviation of each variable, extract abnormal points, judge the trend deviation of pressure and flow rate changes, and generate production abnormal fluctuation results in combination with the abnormal fluctuation time period; Multi-dimensional fault identification module: Based on the abnormal production fluctuation results, a density clustering algorithm is used to analyze key variables, extract key variables in historical data, analyze the change pattern of abnormal fluctuations, combine the historical trends of flow rate and pressure, screen possible fault types, and generate fault pattern identification results; Correlation pattern analysis module: Based on the fault pattern identification results, analyze the equipment operating conditions and geological conditions, extract the correlation of fault patterns, determine the correlation between fault types and different environments, calculate the occurrence frequency of each fault mode, and generate multi-layer correlation pattern analysis results.

2. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The dynamic parameter change mode includes a pressure change rate mode, a flow rate change trend mode, and a temperature fluctuation range; the future production fluctuation trend includes historical production analysis, future production fluctuation calculation, and key data interval screening; the abnormal production fluctuation results include abnormal pressure fluctuations, abnormal flow rate fluctuations, and abnormal fluctuation time periods; the fault mode identification results include pressure fault mode, flow rate fault mode, and historical trend fault type screening.

3. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The downhole dynamic data analysis module includes a pressure analysis submodule, a flow rate analysis submodule, and a temperature analysis submodule, wherein: Pressure analysis submodule: Based on the pressure data of downhole equipment, extract the pressure change value, segment the pressure changes in different time periods, calculate the pressure fluctuation rate and filter the fluctuation range, analyze the change trend of each period, and generate the pressure change pattern; Flow rate analysis submodule: based on the pressure change pattern, extract flow rate data, compare the trend of time and flow rate change, calculate the synchronization of flow rate and pressure fluctuation, screen the time interval with significant flow rate change, compare the fluctuation trend, and generate the flow rate change pattern; Temperature analysis submodule: Based on the flow rate change pattern, extract temperature data, calculate the temperature fluctuation range in each time period, screen the abnormal temperature fluctuation range, combine the flow rate and pressure changes, analyze the correlation between the parameters, and generate a dynamic parameter change pattern.

4. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The digital twin modeling generation module includes a historical data integration submodule, a real-time data processing submodule, and a virtual mapping generation submodule, wherein: Historical data integration submodule: Based on the dynamic parameter change mode, the autoregressive integral moving average model is used to extract the historical production data of the equipment, integrate the data of different production cycles, conduct trend change analysis, summarize the intervals with a long time span, and generate a historical data model; Real-time data processing submodule: based on the historical data model, extract real-time data, compare the real-time changes in each time period, summarize the data fluctuations of real-time pressure, flow rate and temperature in sections, establish the corresponding relationship between the time axis and data changes, and generate a real-time data model; Virtual mapping generation submodule: Based on the real-time data model, historical and real-time data are integrated, the dynamic changes of downhole equipment are virtually mapped, a real-time mapping model of global data is established, and digital twin modeling of oil and gas wells is generated.

5. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The autoregressive integrated moving average model is based on the formula: in: The device production data for the current time point, is a constant term, is the autoregression coefficient, The lag time point Equipment production data, is the sliding average coefficient, The lag time point The error term, is the random error term at the current time point, is the production rate coefficient, is the production rate at the current time point, is the device temperature correction coefficient, The device temperature at the current time point.

6. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The production fluctuation prediction module includes a historical data extraction submodule, a fluctuation trend calculation submodule, and a data interval screening submodule, wherein: Historical data extraction submodule: based on the oil and gas well digital twin modeling, extract the pressure and flow rate data of downhole equipment, analyze the changes in the production cycle, screen the pressure and flow rate values ​​in the key period, integrate the data of different stages and store them by category, and generate a historical production data set; Fluctuation trend calculation submodule: based on the historical production data set, compare the pressure and flow rate changes in different time periods, calculate the amplitude of the fluctuation trend, process the historical data of each period, determine the change trend of pressure and flow rate, combine the real-time data to perform an extended analysis of the historical trend, and generate the fluctuation trend analysis results; Data interval screening submodule: Based on the fluctuation trend analysis results, screen out the time intervals with large pressure and flow rate changes, extract the values ​​of different fluctuation intervals, analyze the fluctuation amplitude and duration, classify and mark the abnormal amplitude intervals and make trend judgments to generate future production fluctuation trends.

7. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The abnormal fluctuation monitoring module includes a prediction deviation calculation submodule, an abnormal point extraction submodule, and a fluctuation trend deviation judgment submodule, wherein: Prediction deviation calculation submodule: Based on the future production fluctuation trend, the actual data of current production pressure and flow rate are extracted, compared with the predicted values, the deviation value in each time period is calculated, the deviation of different variables is classified, and an overall deviation data set is formed to generate a production prediction deviation data set; Abnormal point extraction submodule: based on the production prediction deviation data set, screen out data points with large deviations, extract abnormal changes in pressure and flow rate, classify and summarize different abnormal types, extract the most critical abnormal points in the production process, establish an abnormal data table, and generate a key abnormal data set; Fluctuation trend deviation judgment submodule: Based on the key abnormal data set, the fluctuation trend deviation of pressure and flow rate is judged, and the data of abnormal fluctuation period is combined to analyze the change of deviation value in different time periods, summarize the characteristics of deviation fluctuation, and generate production abnormal fluctuation results.

8. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The multi-dimensional fault identification module includes a key variable analysis submodule, a fluctuation mode screening submodule, and a fault type screening submodule, wherein: Key variable analysis submodule: Based on the abnormal production fluctuation results, a density clustering algorithm is used to extract pressure and flow rate data, analyze the pressure change amplitude in different time periods, calculate the increase and decrease of flow rate, classify related variables, screen out key data points, and generate a key variable set; Fluctuation pattern screening submodule: based on the key variable set, extract the historical data of key variables, analyze the change pattern of each variable in different time periods, screen out the time periods with large abnormal fluctuations, combine the trends of pressure and flow rate, summarize the fluctuation characteristics of each variable, and generate a fluctuation pattern set; Fault type screening submodule: Based on the fluctuation pattern set and combined with the change history of flow rate and pressure, the fluctuation patterns that may cause equipment failure are screened out, the fluctuation types are classified, the type and characteristics of each fault mode are determined, and the fault pattern recognition results are generated.

9. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The density clustering algorithm is based on the formula: in: is the improved local density value, is the number of data points, is the target data point, For the data points, For the Data points With the target data point The Euclidean distance between For the The pressure variation of each data point is For the The flow velocity variation of each data point is For the The length of the time period where the data point is located is other time units, is the Euclidean distance The weight coefficient of is the pressure variation The weight coefficient of is the flow velocity variation The weight coefficient of The length of the time period The weight coefficient of is the bandwidth parameter of the Gaussian kernel function.

10. The oil and gas well digital twin modeling and prediction analysis system according to claim 1, It is characterized in that The correlation mode analysis module includes an equipment operating condition analysis submodule, a fault mode correlation submodule, and a fault frequency calculation submodule, wherein: Equipment operating condition analysis submodule: Based on the fault mode recognition results, extract the operating data of the equipment under different environmental conditions, analyze the changes in flow rate and pressure under various environmental conditions, screen out geological conditions and parameters related to equipment operation, classify the operating performance of the equipment, and generate an equipment operating condition set; Fault mode association submodule: based on the equipment operating condition set, extract the equipment fault data, analyze the correlation between each fault and the environmental conditions, screen out the matching between the fault and the geological environment, analyze the possibility of the fault, and generate the fault mode association result; Fault frequency calculation submodule: Based on the fault mode association results, extract the historical occurrence data of each fault mode, analyze the occurrence frequency of various faults under different conditions, classify the time distribution of fault modes, screen out the high-incidence periods of faults, and generate multi-layer association mode analysis results.

Images